index-constitution 0.6.1

Point-in-time and current constituents for major stock indices (CSI 300, CSI 500, S&P 500, NASDAQ-100), packaged as pandas DataFrames.

Index Constitution

中文

Purpose

This repository was created to make it easier to train quantitative models on major stock indices. Reliable historical index composition data (constituent additions and removals over time) is notoriously hard to obtain — vendors often charge for it, official sources are scattered across PDFs and announcements, and free APIs rarely expose point-in-time membership. Without this data, backtests suffer from survivorship bias and lookahead bias.

This repo collects and normalizes that information into plain CSV files so it can be consumed directly by research and modeling pipelines.

Datasets

Index	Description	Source
CSI 300	Top 300 A-share stocks listed on the Shanghai and Shenzhen exchanges	Official announcements from China Securities Index Co. (csindex.com.cn)
CSI 500	500 mid-cap A-share stocks listed on the Shanghai and Shenzhen exchanges	Official announcements from China Securities Index Co. (csindex.com.cn)
S&P 500	500 leading large-cap U.S. companies listed on U.S. exchanges	Wikipedia: List of S&P 500 companies
NASDAQ-100	100 largest non-financial companies listed on the Nasdaq Stock Market	Wikipedia: NASDAQ-100
Dow Jones Industrial Average	30 large U.S. blue-chip companies in the Dow Jones Industrial Average	Wikipedia: Dow Jones Industrial Average and Wikipedia: Historical components of the Dow Jones Industrial Average

Python package

This repo also ships a small Python library that embeds the CSVs and exposes them as pandas DataFrames.

Install:

pip install index-constitution

Usage:

import index_constitution as ic

ic.list_indices()                    # ['csi300', 'csi500', 'sp500', 'nasdaq100', 'dow30']

ic.latest("sp500")                   # current S&P 500 members
ic.latest("dow30")                   # current Dow 30 members
ic.history("csi300")                 # full CSI 300 history with opt-in/opt-out
ic.constituents_at("sp500", "2015-06-30")   # point-in-time membership
ic.is_member("sp500", "AAPL", "2020-01-02") # True
ic.events("sp500")                   # ticker/name change audit trail
ic.symbol_status("sp500", "ABMD")   # whether a historical symbol is still directly usable

Ticker and name changes

history/*.csv and latest/*.csv use the current canonical ticker and namefor each company across the full membership span. For example, S&P 500 historylists Meta Platforms only as META, even for the period when it traded as FB. The event/us.csv and event/cn.csv files are the audit trail for those changes.

This canonicalization is strongest for pure ticker/name changes. When an event row includes new_symbol, it means this dataset treats the new ticker as the usable successor for historical lookup. For example, FB -> META means you can use META to access the full history for that company in this dataset.

delisting means the old ticker was retired and is no longer directly usable. In that case new_symbol and new_name are left empty because this dataset does not treat any other ticker as its direct successor. For example, ABMD remains valid historical S&P 500 membership data, but the symbol itself stopped trading after Johnson & Johnson acquired Abiomed.

Events are not scoped to a single index — a corporate ticker or name change applies to every index that includes the company. ic.events("sp500") filters the table to events whose old or new symbol ever appeared in S&P 500 history.

is_member() and constituents_at() are strict — they do not resolve old tickers automatically. Use ic.events("sp500") or ic.symbol_status() to tell whether an old symbol maps to a usable successor ticker or is simply delisted.

Use Cases

• Check the current constituents of an index
• Reconstruct point-in-time index membership for backtesting
• Avoid survivorship bias when training quantitative models
• Keep a consistent structure for adding more indices later